A new metric for predicting software change using gene expression programming

Abstract

Software metrics help in determining the quality of a software product. They can be used for continuous inspection of a software to assist software developers in improving its quality. We can also use metrics to develop quality models which predict important quality attributes like change proneness. Determination of change prone classes in an Object-Oriented software will help software developers to focus their time and resources on the weak portions of the software. In this paper, we validate the Chidamber and Kemerer metric suite for building an efficient software quality model which predict change prone classes with the help of Gene Expression Programming. The model is developed using an open source software. We further propose a new metric which can be used for identifying change prone classes in the early phases of software development life cycle. The proposed metric is validated on another open source software and the results show that it can be effectively used by the software industry to classify change prone classes. Identification of change prone classes may help in efficient refactoring and rigorous testing of these classes in the forthcoming releases of the software product.